Gianluigi's profile - activity

As title, I am trying to build new OpenCV 2.4.3 rc with cmake, but this package name is really confused me. What is this package? Should I build it?

I try to build the sample code in tutorial Similarity check (PNSR and SSIM) on the GPU , and there are many error like gpu-basics-similarity.cpp 328 error: no match for ‘operator-=’ in ‘sigma1_2 -= mu1_2’

I try to trace the gpumat.hpp and gpu.hpp, it seems like there is really no implement of operator like +, -, *, +=, -=, ...

Is this a bug or something I misunderstand?

I'm trying to build new OpenCV 2.4.7 with IPP.
If I use old IPP 7.1update1 or uncheck WITH_IPP, it can be build correctly.
But if I upgradde to IPP 8.0 or 8.0update1, it will appear linker error like below:

../../lib/libopencv_core.so.2.4.7: undefined reference to `_intel_fast_memcpy'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__svml_cos4'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__svml_sin2'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__intel_ssse3_rep_memcpy'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__svml_sincos2'
../../lib/libopencv_core.so.2.4.7: undefined reference to `_intel_fast_memset'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__svml_sin4'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__libm_sse2_sincos'
../../lib/libopencv_core.so.2.4.7: undefined reference to `__intel_ssse3_memcpy'

Please help me to correct it, thanks.

Suplement: After I use my correctly built opencv(2.4.7 with IPP 7.1) to build my code. It will appear linker error again, like below:

/usr/bin/ld: cannot find -llibippcore_l.a
/usr/bin/ld: cannot find -llibipps_l.a
/usr/bin/ld: cannot find -llibippi_l.a
/usr/bin/ld: cannot find -llibippcv_l.a
/usr/bin/ld: cannot find -llibippcc_l.a
/usr/bin/ld: cannot find -llibippvm_l.a
/usr/bin/ld: cannot find -ltbb

and these link is come from opencv.pc.
I don't know my link direcctory should be set to /opt/intel/ipp/lib/intel64 or not but I can comfirm the name -llibippcore_l.a is not correct, it should be -lippcore_l.a normally.
The opencv 2.6.1 don't use these linking and can work correctly.

I found the solution, just modify line 39
target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_GPU_SAMPLES_REQUIRED_DEPS})
to
target_link_libraries(${the_target} ${OPENCV_LINKER_LIBS} ${OPENCV_GPU_SAMPLES_REQUIRED_DEPS} cuda)
in "opencv-2.4.6.1/samples/gpu/CMakeLists.txt" can easily fix this problem

Thanks, I've reported it

I add "-L/usr/local/cuda/lib64 -lcuda" to "opencv-2.4.6.1/build/samples/gpu/CMakeFiles/example_gpu_driver_api_multi.dir/link.txt" and it can be build correct, but next target gpu-example-driver_api_stereo_multi will still failed. I think there should be a way to modify CMakeLists.txt to add this to every target, but I have no idea about cmake

I choose another way, I comment "add_subdirectory(gpu)" in "opencv-2.4.6.1/samples/CMakeLists.txt" than it can be build successfully

No, when make detect error it will just leave... Is there a way to build sample only?

I am trying to use ubuntu 12.04 x64 with CUDA 5.5 Production Release to build OpenCV 2.4.6.1
I can build without sample correctly but when I choose to build sample it will failed with

Linking CXX executable ../../bin/gpu-example-driver_api_stereo_multi
CMakeFiles/example_gpu_driver_api_multi.dir/driver_api_multi.cpp.o: In function `destroyContexts()':
driver_api_multi.cpp:(.text._Z15destroyContextsv+0xc): undefined reference to `cuCtxDestroy_v2'
driver_api_multi.cpp:(.text._Z15destroyContextsv+0x1c): undefined reference to `cuCtxDestroy_v2'
CMakeFiles/example_gpu_driver_api_multi.dir/driver_api_multi.cpp.o: In function `Worker::operator()(int) const':
driver_api_multi.cpp:(.text._ZNK6WorkerclEi+0x19): undefined reference to `cuCtxPushCurrent_v2'
driver_api_multi.cpp:(.text._ZNK6WorkerclEi+0x6bf): undefined reference to `cuCtxPopCurrent_v2'
CMakeFiles/example_gpu_driver_api_multi.dir/driver_api_multi.cpp.o: In function `main':
driver_api_multi.cpp:(.text.startup.main+0x94): undefined reference to `cuInit'
driver_api_multi.cpp:(.text.startup.main+0xb1): undefined reference to `cuDeviceGet'
driver_api_multi.cpp:(.text.startup.main+0xd2): undefined reference to `cuCtxCreate_v2'
driver_api_multi.cpp:(.text.startup.main+0xed): undefined reference to `cuCtxPopCurrent_v2'
driver_api_multi.cpp:(.text.startup.main+0x10d): undefined reference to `cuDeviceGet'
driver_api_multi.cpp:(.text.startup.main+0x12e): undefined reference to `cuCtxCreate_v2'
driver_api_multi.cpp:(.text.startup.main+0x149): undefined reference to `cuCtxPopCurrent_v2'
collect2: ld returned 1 exit status
make[2]: *** [bin/gpu-example-driver_api_multi] Error 1
make[1]: *** [samples/gpu/CMakeFiles/example_gpu_driver_api_multi.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
CMakeFiles/example_gpu_driver_api_stereo_multi.dir/driver_api_stereo_multi.cpp.o: In function `destroyContexts()':
driver_api_stereo_multi.cpp:(.text._Z15destroyContextsv+0xc): undefined reference to `cuCtxDestroy_v2'
driver_api_stereo_multi.cpp:(.text._Z15destroyContextsv+0x1c): undefined reference to `cuCtxDestroy_v2'
CMakeFiles/example_gpu_driver_api_stereo_multi.dir/driver_api_stereo_multi.cpp.o: In function `contextOff()':
driver_api_stereo_multi.cpp:(.text._Z10contextOffv[contextOff()]+0xa): undefined reference to `cuCtxPopCurrent_v2'
CMakeFiles/example_gpu_driver_api_stereo_multi.dir/driver_api_stereo_multi.cpp.o: In function `contextOn(int)':
driver_api_stereo_multi.cpp:(.text._Z9contextOni[contextOn(int)]+0x10): undefined reference to `cuCtxPushCurrent_v2'
CMakeFiles/example_gpu_driver_api_stereo_multi.dir/driver_api_stereo_multi.cpp.o: In function `Worker::operator()(int) const':
driver_api_stereo_multi.cpp:(.text._ZNK6WorkerclEi+0x14): undefined reference to `cuCtxPushCurrent_v2'
driver_api_stereo_multi.cpp:(.text._ZNK6WorkerclEi+0x106): undefined reference to `cuCtxPopCurrent_v2'
CMakeFiles/example_gpu_driver_api_stereo_multi.dir/driver_api_stereo_multi.cpp.o: In function `main':
driver_api_stereo_multi.cpp:(.text.startup.main+0x80b): undefined reference to `cuInit'
driver_api_stereo_multi.cpp:(.text.startup.main+0x822): undefined reference to `cuDeviceGet'
driver_api_stereo_multi.cpp:(.text.startup.main+0x83d): undefined reference to `cuCtxCreate_v2'
driver_api_stereo_multi.cpp:(.text.startup.main+0x85c): undefined reference to `cuDeviceGet'
driver_api_stereo_multi.cpp:(.text.startup.main+0x877): undefined reference to `cuCtxCreate_v2'
collect2: ld returned 1 exit status

It seems like forgetting link libcudart, I think this can be fixed by modifying CMakeLists.txt but I don't know how to fix it.

I use QT4 and cmake-gui, although it will appear qt5 error but it can still configure, generate, and make correctly.

Sorry, it can configure now, but it appear many error when I build it. Like, error: calling a __device__ function("AddMat") from a __host__ function("addMat") is not allowed detected during instantiation of "void arithm::addMat<T,D>(cv::gpu::PtrStepSzb, cv::gpu::PtrStepSzb, cv::gpu::PtrStepSzb, cv::gpu::PtrStepb, cudaStream_t) [with T=ushort, D=double]"

Thanks, it works!

The release package 2.4.5 can be built with CUDA 5.0 using cmake,

When I use cmake to configure, it will it cannot found CUDA_npp_LIBRARY

CMake Error: The following variables are used in this project, but they are set to NOTFOUND.
Please set them or make sure they are set and tested correctly in the CMake files:
CUDA_npp_LIBRARY (ADVANCED)
    linked by target "opencv_core" in directory /home/gianluigi/Lib/opencv-2.4.5/modules/core
    linked by target "opencv_gpu" in directory /home/gianluigi/Lib/opencv-2.4.5/modules/gpu
    linked by target "opencv_nonfree" in directory /home/gianluigi/Lib/opencv-2.4.5/modules/nonfree

I think this is caused by NVIDIA split the libnpp to libnppc, libnppi and libnpps in CUDA 5.5. Is there any way to fix this problem?

I just can't find the button anywhere.
The document says it will attach on control panel, but I never saw one.
I use OpenCV 2.4.3 with Qt4 on Ubuntu 10.04.
Below is my short test code.

#include <string>
#include <opencv2/opencv.hpp>

using namespace std;
using namespace cv;

void on_push(int state, void*) {}

int main(int argc, char* argv[])
{
    char winName[] = "TEST";
    Mat image = imread("000001.jpg", 1);

    namedWindow(winName, CV_WINDOW_NORMAL | CV_WINDOW_KEEPRATIO);
    resizeWindow(winName, 1280, 960);
    createButton("draw", on_push);

    imshow(winName, image);
    waitKey();

    return 0;
}

Is there any problem when I use it?

I have a GpuMat Image(H, W, CV_8UC3)
May I bind it using these function?

texture<uchar3, 2, cudaReadModeNormalizedFloat> rgbTex;
cudaChannelFormatDesc desc = cudaCreateChannelDesc<uchar3>();
cudaBindTexture2D(0, rgbTex, Image.data, desc, W, H, Image.step));

Thanks, it's helpful.